[PATCH 0/3] commit: fix abbrev-sha regression

[Tay Ray Chuan <rctay89@gmail.com>]

After c197702 (pretty: Respect --abbrev option), the summary output for
git-commit began to throw up non-abbreviated SHA-1s.

For example,

  $ git commit --allow-empty -m "empty"
  [master f869bf09c8eefee54094bb21387241aaf5f10324] empty

I also notice that this happens for merges after conflict resolutions.

This afflicts 'master', but not any release.

Here's the hunk that fixed it:

-->8--
diff --git a/builtin/commit.c b/builtin/commit.c
index a4e4966..ab4a7cf 100644
--- a/builtin/commit.c
+++ b/builtin/commit.c
@@ -1166,6 +1166,7 @@ static void print_summary(const char *prefix, const unsigned char *sha1)
 		struct pretty_print_context ctx = {0};
 		struct strbuf buf = STRBUF_INIT;
 		ctx.date_mode = DATE_NORMAL;
+		ctx.abbrev = DEFAULT_ABBREV;
 		format_commit_message(commit, format.buf + 7, &buf, &ctx);
 		printf("%s\n", buf.buf);
 		strbuf_release(&buf);
-->8--

However, I noticed that this would also work:

-->8--
diff --git a/builtin/commit.c b/builtin/commit.c
index ab4a7cf..13a30ab 100644
--- a/builtin/commit.c
+++ b/builtin/commit.c
@@ -1148,7 +1148,7 @@ static void print_summary(const char *prefix, const unsigned char *sha1)
        rev.verbose_header = 1;
        rev.show_root_diff = 1;
        get_commit_format(format.buf, &rev);
-       rev.always_show_header = 0;
+       rev.always_show_header = 1;
        rev.diffopt.detect_rename = 1;
        rev.diffopt.rename_limit = 100;
        rev.diffopt.break_opt = 0;
-->8--

This would make the code block that handled the situation in which
log_tree_commit() did not print any output redundant, so perhaps this
would also follow:

-->8--
diff --git a/builtin/commit.c b/builtin/commit.c
index ab4a7cf..e8def55 100644
--- a/builtin/commit.c
+++ b/builtin/commit.c
@@ -1163,13 +1163,7 @@ static void print_summary(const char *prefix, const unsigned char *sha1)
                initial_commit ? " (root-commit)" : "");

        if (!log_tree_commit(&rev, commit)) {
-               struct pretty_print_context ctx = {0};
-               struct strbuf buf = STRBUF_INIT;
-               ctx.date_mode = DATE_NORMAL;
-               ctx.abbrev = DEFAULT_ABBREV;
-               format_commit_message(commit, format.buf + 7, &buf, &ctx);
-               printf("%s\n", buf.buf);
-               strbuf_release(&buf);
+               die("unable to print summary");
        }
        strbuf_release(&format);
 }
-->8--

In a sense, it reverts parts of bf82a15 (commit: do not add extra LF at
the end of the summary.), except for the extra LF. Would this be a more
sensible route?

Contents:
[PATCH 0/3] commit: fix abbrev-sha regression
[PATCH 1/3] t7502-commit: add tests for summary output
[PATCH 2/3] t7502-commit: add summary output tests for empty and merge commits
[PATCH 3/3] commit: show abbreviated sha for commits with empty diffs